app.py

import gradio as gr
import json
import subprocess

def run_pipeline(pdf_file, system_prompt, max_step, learning_rate, epochs, model_name):
    # Construct job input
    data = {
        "input": {
            "pdf_file": pdf_file,
            "system_prompt": system_prompt,
            "max_step": max_step,
            "learning_rate": learning_rate,
            "epochs": epochs,
            "model_name": model_name
        }
    }

    try:
        # Call handler.py using the constructed input
        input_json = json.dumps(data)
        process = subprocess.Popen(
            ['python3', 'handler.py', '--test_input', input_json],
            stdout=subprocess.PIPE,
            stderr=subprocess.PIPE,
            text=True
        )
        stdout, stderr = process.communicate()

        if process.returncode != 0:
            return {"status": "error", "details": f"Handler failed: {stderr.strip()}"}

        # Parse the handler output
        try:
            handler_output = json.loads(stdout)
            if handler_output.get("status") == "success":
                # Extract relevant fields
                model_name = handler_output.get("model_name", "N/A")
                evaluation_results = handler_output.get("evaluation_results", {})
                semantic_score = evaluation_results.get("average_semantic_score", "N/A")
                bleu_score = evaluation_results.get("average_bleu_score", "N/A")

                return {
                    "status": "success",
                    "model_name": model_name,
                    "evaluation_results": {
                        "average_semantic_score": semantic_score,
                        "average_bleu_score": bleu_score
                    },
                }
            else:
                # Return error from handler
                return handler_output

        except json.JSONDecodeError:
            return {"status": "error", "details": f"Invalid JSON from handler: {stdout.strip()}"}

    except FileNotFoundError:
        return {"status": "error", "details": "Handler script not found"}
    except Exception as e:
        return {"status": "error", "details": str(e)}

# Define Gradio interface
with gr.Blocks(css='''
    .gradio-container {
        background-color: #121212;  /* Dark background */
        color: #f1f1f1;             /* Light text color */
        padding: 20px;
        font-family: 'Arial', sans-serif;
    }

    .gr-row {
        margin-bottom: 20px;
    }

    /* Styling for Textboxes and Numbers */
    input[type="text"], input[type="number"], textarea {
        background-color: #f0f0f0;  /* Light grey background for inputs */
        border: 1px solid #ccc;     /* Light grey border */
        color: #000;                /* Black text inside the inputs */
        border-radius: 8px;
        padding: 10px;
        font-size: 16px;
        width: 100%;
        box-sizing: border-box;
    }

    /* Styling specific to textarea placeholder */
    textarea::placeholder {
        color: #999;  /* Slightly darker grey placeholder text */
    }

    /* Button styling */
    button {
        background-color: #4CAF50; /* Green button */
        color: white;
        border: none;
        padding: 12px 20px;
        cursor: pointer;
        font-weight: bold;
        font-size: 16px;
        transition: background-color 0.3s ease;
        border-radius: 8px;
    }

    button:hover {
        background-color: #3e8e41;  /* Darker green hover effect */
    }

    /* Styling for JSON output */
    .gr-json {
        background-color: #333;     /* Dark background for JSON output */
        border: 1px solid #444;     /* Slightly lighter border */
        padding: 12px;
        font-size: 14px;
        max-height: 300px;
        overflow-y: auto;
        margin-top: 10px;
        color: #f1f1f1;              /* Light text color */
    }

    /* Adjust margins for all inputs */
    .gr-row .gr-textbox, .gr-row .gr-number {
        margin-bottom: 15px;
    }
''') as demo:

    # Add Heading at the top
    gr.Markdown(
        '<h2 style="color: #87CEEB; text-align: center;">🤖 Fine-tuning Pipeline Configurator</h2>'
    )

    # Layout structure with improved spacing
    with gr.Row():
        with gr.Column(scale=2):
            pdf_file = gr.Textbox(label="PDF Blob Name", placeholder="Enter the PDF file", value="finetuning_pipeline_testing_V1")
        with gr.Column(scale=3):
            system_prompt = gr.Textbox(label="System Prompt", placeholder="Enter system instructions or context", value="You are a helpful assistant that provides detailed information based on the provided text.")
        with gr.Column(scale=2):
            max_step = gr.Number(label="Max Steps", value=150)
        with gr.Column(scale=2):
            learning_rate = gr.Number(label="Learning Rate", value=2e-4)
        with gr.Column(scale=2):
            epochs = gr.Number(label="Epochs", value=10)
        with gr.Column(scale=3):
            model_name = gr.Textbox(label="Model Name", placeholder="Enter the model name")

    result_output = gr.JSON(label="Pipeline Results")
    run_button = gr.Button("Run Pipeline")

    # Trigger the function when the button is clicked
    run_button.click(
        run_pipeline,
        inputs=[pdf_file, system_prompt, max_step, learning_rate, epochs, model_name],
        outputs=[result_output]
    )

# Run Gradio app
demo.launch()